Abstract:This paper introduces a novel method for image colorization that utilizes a color transformer and generative adversarial networks (GANs) to address the challenge of generating visually appealing colorized images. Conventional approaches often struggle with capturing long-range dependencies and producing realistic colorizations. The proposed method integrates a transformer architecture to capture global information and a GAN framework to improve visual quality. In this study, a color encoder that utilizes a random normal distribution to generate color features is applied. These features are then integrated with grayscale image features to enhance the overall representation of the images. Our method demonstrates superior performance compared with existing approaches by utilizing the capacity of the transformer, which can capture long-range dependencies and generate a realistic colorization of the GAN. Experimental results show that the proposed network significantly outperforms other state-of-the-art colorization techniques, highlighting its potential for image colorization. This research opens new possibilities for precise and visually compelling image colorization in domains such as digital restoration and historical image analysis.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is the challenges faced by existing image coloring methods in generating visually appealing color images. Specifically, traditional methods have difficulties in capturing long - distance dependencies and generating realistic colors. To address these issues, the authors propose a new image coloring method, which utilizes color transformers and generative adversarial networks (GANs) to improve the quality of image coloring. ### Main Problem Summary: 1. **Capturing Long - Distance Dependencies**: Traditional methods are difficult to effectively capture global information in images, resulting in a lack of coherence and consistency in the generated color images. 2. **Generating Realistic Color Images**: Existing automatic coloring techniques are prone to color distortion, over - saturation, or color bleeding when dealing with complex scenes. 3. **Balancing Details and Global Understanding**: During the coloring process, how to understand the broader context while maintaining fine - grained textures is a difficult point. ### Proposed Method: To overcome the above challenges, the authors propose a new method that combines a color encoder, a color transformer, and a generative adversarial network (GAN) based on an encoder - decoder architecture. Specific improvements include: - **Color Encoder**: Generate color features using a random normal distribution and fuse them with grayscale image features to enhance the overall representation of the image. - **Color Transformer**: Capture global information through a transformer architecture to ensure that color assignment is more context - compliant and coherent. - **Generative Adversarial Network (GAN)**: Utilize the adversarial training mechanism to improve the visual quality of the generated image, making the generated color image more realistic. ### Experimental Results: The experimental results show that this method significantly outperforms the existing state - of - the - art coloring techniques in multiple evaluation metrics such as PSNR, SSIM, and color richness. Especially in terms of naturalness and closeness to real colors, this method performs excellently and can generate visually more appealing and realistic color images. ### Formula Presentation: 1. **Formula for the Color Transformer Module**: \[ x_i=\text{Concat}(x_e, x_{ce})\quad(1) \] \[ x_c = \text{Conv}(x_i)\quad(2) \] \[ x_1=T_w^1(x_c)\quad(3) \] \[ x_2=T_w^2(x_1)\quad(4) \] \[ x = x_c + x_2\quad(5) \] 2. **Objective Function**: \[ \mathcal{L}=\lambda_g\mathcal{L}_g+\lambda_p\mathcal{L}_p+\lambda_{L1}\mathcal{L}_{L1}+\lambda_c\mathcal{L}_c\quad(6) \] - \(\mathcal{L}_g\) is the adversarial loss (Wasserstein GAN loss) - \(\mathcal{L}_p\) is the perceptual loss - \(\mathcal{L}_{L1}\) is the \(L1\) loss - \(\mathcal{L}_c\) is the color loss Through these improvements, this method not only improves the quality of image coloring but also provides new possibilities for digital restoration, historical image analysis, and other fields.

Transforming Color: A Novel Image Colorization Method

Transforming Color: A Novel Image Colorization Method

Towards Vivid and Diverse Image Colorization with Generative Color Prior

CT2: Colorization Transformer via Color Tokens

SPDGAN: A Generative Adversarial Network based on SPD Manifold Learning for Automatic Image Colorization

Double-Channel Guided Generative Adversarial Network for Image Colorization

$\mathrm {CT^2}$: Colorization Transformer via Color Tokens.

Using colorization as a tool for automatic makeup suggestion

Colorful Image Colorization

Style Transformer for Image Inversion and Editing

DDColor: Towards Photo-Realistic Image Colorization via Dual Decoders

Two-stage image colorization via color codebook

Globally and Locally Semantic Colorization via Exemplar-Based Broad-GAN

Grayscale Image Colorization with GAN and CycleGAN in Different Image Domain

Colourization of Black and White Images using Deep Learning

ChromaGAN: Adversarial Picture Colorization with Semantic Class Distribution

SCGAN: Saliency Map-guided Colorization with Generative Adversarial Network

UGSC‐GAN: User‐guided sketch colorization with deep convolution generative adversarial networks

A Novel Color Image Enhancement Method by the Transformation of Color Images to 2-D Grayscale Images

Rethinking low-light enhancement via Transformer-GAN

ColorUNet: A convolutional classification approach to colorization